Hello, everyone. I am thrilled to be here to talk to all of you today about how do I get it all done in a day, ICU pharmacist practice models, and balancing the work. By the end of this presentation, I want us all to be able to characterize the relationships of ICU clinician workload, patient-centered outcomes, as well as clinician well-being, discuss how ICU pharmacist practice models integrate into interprofessional team-based care, and finally, ideate on best practices of ICU pharmacist practice models. So we'll begin with the first section, which is going to be a discussion of workload and patient outcomes. Put simply, there is a relationship between clinician workload and the quality of care that they provide. And then there is a relationship between the quality of care that they provide and patient-centered outcomes. So overall, we have a relationship between how much we ask someone to do in a given day or shift, and the quality of care and patient outcomes. And this is really important for us to think about, because I kind of like to joke, probably everyone on this call has at least done one journal club on should we be using vitamin C and sepsis. And that's a nice thing to talk about, I have nothing against looking at the latest literature. But I look at the vitamin C impact, and then I look at something like for pharmacists, the fact that having a pharmacist on interprofessional rounds is associated with a reduction in the odds of mortality by 22%, and I think sometimes we're losing the forest for the trees. So what happens with overwork or a high workload? There's a couple of things that can happen with a non-optimized shift. One thing is that if you have too many items to do in a day, by the end of your shift, you just stay past your shift. And you can say, hey, a couple times you get busy, it's right before the holidays, something like that, you end up with one long shift. But when you start adding one long shift after another long shift over the course of a year, two years, three years, a decade, that can really add up to things like burnout and attrition from the field. The other thing that can happen is that we don't do other things that are important in our job. So generally speaking, we're all going to prioritize taking the best care of patients that we can. So I have in the lighter blue direct patient care responsibilities. So if that is a large amount and it gets larger and it's not optimized, what it's going to cut into is indirect patient care activities. So you can say, well, that's good, we're taking care of patients, right? But sometimes the indirect patient care activities can have just as much of a profound effect, if not more of a profound effect, than direct patient care things. So an example that I have is at one point in time, we didn't have a good process for using neuromuscular blockade in my institution. And so I finally sat down and made the protocol, the order set, all the different things, which took a while. And it took me from having to deal with that many days to never having to deal with it again. And moreover, that even on the days that I wasn't there, it was still being done correctly. And even after I left for a faculty position, it's still being used and still being done correctly. And those are nice things to realize, but they did take time and effort to do. So what happens is that you have people that know that if they had more time, they would be reading the latest literature, they would be updating that order set that they know needs to be updated, they would be spending more time with their students. And all of those things can create a really unpleasant workday, not to mention they can have direct impact on patient outcomes. So workload can be optimized for key stakeholders, for patients, clinicians, and institutions. It shouldn't just be that an institutional budget is dominating the workload that we have, and nor should it be any of these individuals, because the reality is we all three kind of need each other in order to have the best possible outcome. So certainly an institution might think that it's cost effective to overwork their people, but if they have lots and lots of turnover and burnout, we know that that is costly to an institution to have to, you know, constantly doing orientations and trainings, that's a known thing. But we also know that burnout is related to the quality of patient care that's provided, which is directly related to outcomes, which also affects both the patients as well as the institutions and quality markers. So all of these things are interrelated. So somewhere there is, and this is kind of a theoretical construct, but there is somewhere a sustainable amount of workload for a clinician that provides optimal patient care and is reasonable towards institutional budgetary constraints and things like that. And so our goal right now is to find that sweet spot. So there is a precedent for evaluating ICU clinician workload as a release to outcomes, particularly with our nursing and physician colleagues. And our nursing colleagues have really done a fantastic job showing that mortality rates increase with the bed-to-nurse ratio when that increased as well. So as their workload increased, we saw worse things happening. And another key study showed that the intensivist ICU bed ratio of being over 1 to 15, while it didn't have an observed effect on ICU mortality, it did show a prolonged length of ICU stay. And ICU stay is surely a patient-centered outcome because every day you're longer in the ICU, one, it's just unpleasant, and two, it's related to increased risk of complications. But it's also an important concept from a financial perspective. It's directly related to costs. So when you start seeing these things, I think this is where you can say there is a relationship between important outcomes and the workload. For pharmacists, we have a plethora of what I call all-or-nothing studies, basically where there was no pharmacist services, and then we added pharmacist services, and we saw some improvement, be it in medication adherence or patient-centered outcome or some type of process-related outcome. Generally, pharmacists are associated with good things. And so my argument here is that there is really an overwhelming body of literature that supports that pharmacists improve outcomes. So sometimes, at least for me, when I open up the latest pharmacy journal and I see another study of can pharmacists improve X outcome, I get a little frustrated because I think, of course, we can improve that outcome. We're trained, we have a four-year doctorate degree in clinical care. We often have two years of residency training, often go on for board certification and do other types of advanced things. So there's no doubt in my mind that a pharmacist can improve an outcome. And so what I like to say is that if there is a patient and there is a drug involved, we know a pharmacist improves the outcomes in some way. And so what I would love is that for those studies, as opposed to answering the question of should pharmacists be there, because the answer is yes, they should, is to say this was our process for the pharmacist being there. This is how we implemented this to be sustainable, integrated into the workload. This is our protocol. We're publishing it so you can copy and paste it and use it at your institution. And far more focus on the nuts and bolts of how that happened, not telling me that a pharmacist improved an outcome, because we know that that's going to be true. But importantly, our focus has been so much so on this kind of justification of positions and justification that pharmacists should be involved in services, we really haven't spent a lot of time looking at workload and how does workload relate to outcomes. So I'm going to go through a few of the key studies that we've got that to me kind of say that this relationship does exist, but also probably serves as an impetus that we need to be doing more studies like this. This was a meta-analysis completed by Lee and colleagues in 2019, and it evaluated 14 different articles assessing the role or the impact of critical care pharmacists on the multidisciplinary team, basically en route. And the key parts that they found is that there was a reduced likelihood of mortality with an odds ratio of 0.78. There was a reduction in ICU length of stay for 1.3 days, and reduction in adverse drug events by 70%. And these are really, to me, really big, really important numbers. If you want to consider it the mic drop, I don't need another study saying that pharmacists on rounds are important, but I think this is an incredibly important study for us to have and to understand that that impact size of a decrease in mortality odds by 22%, that's big. And again, that's bigger than a lot of other things that we talk about in the ICU that maybe improve outcomes or maybe not. We sit here and talk about balanced fluids, and balanced fluids doesn't have anywhere close to a mortality reduction like this. So this is a really important study, and I think shows that if you're ever in a situation where there's a pharmacist not on rounds, you're potentially increasing the risk of mortality for those patients. So situations right now of you cross-cover and you're not available to another team because they round at the same time, weekends, situations where there are just uncovered units, all of those things, to me, are significant areas for improvement, or potentially different teams. Again, all of that, I think, should be a corollary to when you read this study to realize that there is a team rounding without an ICU pharmacist, that's a problem. This was a really neat study that came as an ancillary study to the PharmCRIT study that evaluated what happened when you had multiple rounding teams and what happened with pharmacist interventions. So what they found was that when a pharmacist was responsible for multiple rounding teams, it decreased the pharmacist's efficacy. They were less likely to have their interventions accepted. I would even go so far as to say that not only were they likely not accepted, I would also wonder if overall there were more interventions that were never attempted in these situations because the pharmacist was busy. That's obviously a step beyond what this study shows, but I do think that it checks out with our experiences, or certainly my experience. When I am cross-covering onto a team and I show up and it's like, I think that we should discontinue these antibiotics, but I haven't been there for rounds, I don't know the whole story. It's hard for me to say it with as much confidence as if I was standing there with the team the whole time. It's like, yeah, this makes total sense. So I think this is something that we all know anecdotally, but I think it's really important to see that this was true. I also think it's important to see that this is true because this PharmCRIT study is the same data set that we're going to talk about in a couple of slides that related workload to quality. The PharmCRIT study was a very large study with 215 ICU pharmacists, 85 centers, with 55,000 interventions recorded over the course of 27,000 adult patient days. So just a very, very large undertaking, and congrats to those investigators because this was a very important study that needed to happen. PharmCRIT found that there was a return on investment of about $3 to $9 per every dollar spent on a pharmacist's salary, and that is just an incredible number to see. So again, it's a little bit difficult. Sometimes cost avoidance is not a perfect metric because sometimes the discussion is that the numbers can be a little bit inflated, and that it's hard to maybe see a 10 to 1 ROI. However, there was another study that was recently published about evening pharmacist services and cost avoidance that used extremely conservative numbers and still saw an ROI in this range. So I am certainly a believer that adding an extra pharmacist to your team, you're going to see a really important and meaningful return on investment with regard to the types of things that they're doing. Another study related to the PharmCRIT study was an evaluation of the MRCICU score, trying to understand and validate it in a large population. So while the study did show that MRCICU was related to, as it increased, it increased mortality risk, length of stay, as well as related to pharmacist's workload. As it increased, so did the number and intensity of pharmacist's interventions. But a really important finding was that as the pharmacist to patient ratio went up, we saw an increase in length of stay. So essentially what we're saying here is that as the workload increased for a pharmacist, so too did the patient's length of stay. And on top of that, we also saw that the number of interventions was reduced. And so there is a very interesting story potentially being shown here, that among patients with the same amount of medication regimen complexity, so the same MRCICU score, so they have the same type of medication, same complexity that we're working with, when the pharmacist had more patients to take care of, they could not intervene as much, so total quantity, but that also could not intervene at the same quality as measured by intensity score. And that there was, again, a relationship, too, of length of stay, of increase in length of stay. So while the study was not originally designed to see that, and the relationships are, the signal is definitely present, I'm not going to say it's super strong, I think this was a really important finding that, again, speaks to the story of as you increase workload, you're decreasing the quality of care, which relates, again, to patient outcomes. I do also want to put a plug in here. If you've noticed, the last three slides are all based on the PharmCRIT study, which was a very large undertaking by the critical care pharmacist community in order to have robust enough and large enough data that we can make some different findings like this, and I think this speaks more, too, to where we need to be going over the next 5, 10, 15 years of more, you know, basically the community of critical care pharmacists getting together to answer important questions like this. So I just put that in your mind of, you know, how can we do more things of a meaningful scale like this. This study was a cross-sectional survey of critical care pharmacists with about 185 respondents, and the intent was to kind of identify a baseline pharmacist to patient ratio of those respondents, but also to see if there was any relationship to the perceptions of care, and it was found that higher workload did reduce the pharmacist's perception of the quality of care that was provided. So as the number of patients went up, the number of people that felt that they were providing high-quality care went down, and I think this is important for two reasons. One, I trust a critical care pharmacist's assessment. If they don't think they did a good job on something, I believe that, and I think that that is an important thing to realize that if you have a, you know, a highly qualified individual, a professional, saying this wasn't my best work, that's something we should listen to. The other part I think that's important is that if you feel that you're not doing a good job at your work, that you are not having good efficacy, that actually is one of the pillars of burnout, that you feel that you are unable to do a good job. And so to me, I have wondered if some of the reason that we struggle with burnout in the field is that we are, we're setting people up to fail from the outset, and it's important that we think through how can we help someone who is truly dedicated to helping others and into doing an excellent job, allowing them to have the space to do an excellent job. And kind of as an aside, this ability to motivate and support highly driven individuals, which again I would say critical care pharmacists fall into a highly driven category, that's a classic organizational psychology concept that if you have a highly motivated individual, what you need to do is give them the tools and the support and the infrastructure to be successful, and they will do the rest. So if we're not doing that, it's hard for an individual to overcome those things. This was another cross-sectional survey conducted of critical care pharmacists that evaluated the relationship, or looked at burnout in particular. But it was also curious of how burnout related to the position paper in terms of fundamental versus desirable categories and the different types of patient care versus indirect patient care that they were providing. And so important kind of takeaways from this study, first, over half of the respondents of the 211 respondents reported burnout, and that's a large number, and I think worthy of note. And this I think is also interesting, of the respondents. So this is still someone who had the bandwidth and the ability within them to answer a survey purely out of the goodness of their hearts for a research team that answered that. So I have to think that the other people that weren't responding were likely burnt out. So I think 50% is likely under-reporting. But it was also really interesting to see that the performance of optimal activities was associated with reduced burnout. And so the question is, okay, what are these types of optimal activities? Well, it was things like professional development type stuff, like publishing in the research or providing peer review, but it was also things like indirect patient care, you know, providing quality improvement to your institution. And again, I think of, you know, a highly driven individual who wants to see best outcomes, if they go into the hospital every day and look at a problem and just know they can barely get through the day and can't solve that problem, I think day in and day out we're going to see problems with burnout eventually. And so the fact that we can, you know, when we think about how we're going to design a pharmacist position, we need to honor and support the fact that critical care pharmacists enjoy having the space and the time to do some other activities that aren't directly patient care, but that again directly support patient care. You know, the fact that you are, you know, providing a peer review in the literature and doing a good job of that, you're learning. You're learning things that are going to influence your practice. The fact that you're, you know, doing a medication use evaluation, again, improving the care that we provide. And I think that all of those things are related, that when you have a chance to be successful, you are also less likely to be burned out. So the key takeaway of this section is that a really important next step to improving ICU patient outcomes is optimizing clinician workload. And that when we maximize those clinicians in the context of a team, that that is going to improve outcomes. And here I mean, you know, the presence of a pharmacist on interprofessional routes. And I think that this type of approach, this type of literature likely has ramifications across the ICU care team. You know, just because there's not a study right now of dieticians or occupational therapists in the ICU and their workload doesn't mean that those things aren't incredibly important as well. So I think looking at workload, that really is a next step for critical care as a whole. So the next section we're going to discuss is how pharmacists are integrated into team-based care and the importance of medication management in particular. A really important concept, which I don't think is going to come as a shock to a critical care pharmacist, is that medications are independent risk factors for positive outcomes, as well as ICU complications. So they are causative agents. An interesting reality, though, is that many times when we look at prediction modeling type questions, medications are ignored. I think they're ignored because they are complex. But I do think that when we're discussing, you know, why does medication management matter, it matters because medications, again, they're independent causative agents towards outcomes and complications. So here I have these various scenarios. And scenario two is, you know, you give a drug and there's only harm versus give a drug and there's only benefit and you have a little smiley face here. But scenario one is far more what we're dealing with, which is that every medication we give is associated with both benefits and harms, and we're trying to balance those things. Again, although this may appear obvious, this type of kind of causal pathway has important ramifications when we start getting into prediction modeling, causal inference, as well as just discussing why medication management matters. The direct kind of implication of that causal pathway is that pharmacists are trying to maximize benefits and reduce harms of ICU medications. So in this little pathway, you have a patient, and then you have a pharmacist providing an intervention on a drug therapy, which is then ideally reducing the ADE, reducing the complication, and then improving a patient-centered outcome. Again, this kind of seems obvious. It's what you guys are doing all day, every day. But this is actually relatively novel as a concept. You know, so if you've read Wes Ely's, you know, Every Deep Drawn Breath, you know, we were just almost wantonly using medications with almost no thought or concept to the fact that they could potentially be hurting a patient and hurting far beyond the benefit. But I think that this kind of newfound respect for what medications can do is, again, something that we're really almost just starting to, like, appreciate, and it's also more reason that having a pharmacist present is important. But I think that some of this kind of, this concept explains why this wasn't always the way we were doing things. It's because I don't think we necessarily truly respected the power of medications for good and ill. So the question becomes, how can we quantify medications as independent risk factors for outcomes as well as workload prediction? There are some challenges to this. One of the challenges is that the average ICU patient has 13 medications, oftentimes over 20, at any given time. So that's a lot of variables to add into a regression or even a machine learning model for that matter. The other issue is that unlike lactate, which is, one, a single value, and really it'd even be a binary value of, you know, it's less than or over some threshold, is that medications, you really need to know the name, the dose, the frequency, the formulation, as well as patient information to interpret it. So if I tell you a patient's on cefepime 2Q12, you're going to say, okay, and you have no idea if that is they have renal failure and that dose is four times too high, or if they're on CRT and it's actually maybe two grams too low, or anything else for that matter. So you need to have a lot of information to interpret whether it's correct or not. Finally, the quote unquote ground truth is difficult to identify. What is the most correct medication regimen? So for something like, I don't know, having an MI and going to a cath lab, you know, the ground truth, generally speaking, is if you have a myocardial infarction, you need to end up in the cath lab having something happen to you. There's not a lot of situations where that's going to not be the answer. And so it's fairly easy for us to say, this was correct, this was incorrect, as terms of a course of action. But to say that cefepime versus piperacillin-tazobactam, which is more correct, well obviously we could have a whole two-hour discussion on that. You know, should we be using vancomycin or lenazolid? Again, this could be a very, you know, sometimes there's a really obvious answer. If they had VRE, okay, then maybe not vancomycin, but what happens if they didn't? And so those type of nuances can make understanding ground truth quite difficult. And so these are the reasons why, although we are maybe aware of this type of causal inference pathway, we're not always thinking in terms of how, what are medications doing in relation to these, you know, overall outcomes. So given the challenges of including medications into this type of prediction modeling, the question is, you know, can medication regimen complexity maybe serve as a tool that summarizes the importance of medications in the ICU? Can we use that to predict patient outcomes? Can we use that to predict pharmacist workload? Is this something that can help us, you know, optimize the type of workload that we're giving? So the purpose of this score is that it can objectively and reproducibly quantitate the complexity of the medication regimen. So I have a picture of it over here, so you know, digoxin's worth three points, vancomycin's worth three points, they're on a bowel regimen, that's a point. And the idea is that my patient is going to have a standardized medication regimen complexity that you can trust and compare to your patients. The process that the MRCIC went through has included convergent, divergent, test, retest, internal validity, external validity, and so forth, with the goal that this is something that we could truly create a common language with. It has established correlations with pharmacist workload, including number of interventions, the intensity of those interventions, number of orders verified, as well as number of IVINs placed in the EPIC system as part of routine patient care. It's got correlations to ICU complications, including fluid overload, prolonged mechanical ventilation, and it also is related to patient-centered outcomes, including mortality and length of stay. So the question is, can a tool like this be used in, you know, artificial intelligence or prediction-type modeling? We have found that medication data can improve mortality prediction, particularly when used with artificial intelligence methods. So this was a really exciting paper, because essentially we took a look at whether or not adding medication data to severity of illness data, specifically the Apache score, would improve mortality prediction, and it did. And also when we used different types of machine learning methods, we found that the AUROC curve was improved with some of these different methods. And so while this had a relatively small sample size, I think it does open up the hypothesis that we should be including medications in more prediction modeling, and also that machine learning may offer us some unique insights. Our team also conducted a study evaluating if medication data improved mortality prediction using more traditional regression modeling, and importantly included severity of illness data, including SOFA and Apache score, in a slightly larger cohort of patients. So this had 991 patients. And again, when we included MRC ICU score plus SOFA plus Apache, we had the highest AUROC curve. And I think that this is a really interesting thing to realize, that potentially we should have indicators for patients' needs and severity that includes this type of data within it. One of the more interesting findings that we had with this study was the possibility that there is kind of a Goldilocks zone for medication regimen complexity and mortality. So originally we had assumed more of a linear relationship. As MRC score goes up, so too would mortality, because it's an indicator of overall severity of illness. But there's a possibility that certain patients are very, very sick, which means they do have a high risk of mortality, but if you give them the correct medications and the correct medication complexity, that they do better. Think about a septic patient that receives the appropriate broad-spectrum antibiotics. But conversely, a patient that does not need a bunch of medication intervention that does get that for whatever reason, potentially that is increasing their risk. So we actually have to match the complexity to the mortality risk. So more of a non-linear relationship that is potentially better modeled with either a spline or some type of other type of non-linear type experience, or again, machine learning might be interesting here as well. And again, while I think that this makes intuitive sense for a clinician to say, yeah, of course if we give neuromuscular blockers to a patient that doesn't need them, that's just increasing the harm without benefit. But I think that one of the advantages here is just because maybe we know something or think we know something intuitively doesn't mean we can necessarily see it with mathematics. And the idea and the concept of having mathematics and big data to help drive this is, one, we can maybe see new insights that we didn't know of before, or two, that it may help us with prediction in terms of probability monitoring. I want to give this therapy and it has X risk of harm versus Y risk of benefit, and having that type of mathematics to support your decision-making may improve a clinician's heuristic sense. We have observed that fluid overload prediction may be enhanced with the MRC score, but particularly when machine learning is used. So what was really interesting in this paper, which has recently been accepted to scientific reports, was that when we set up our prediction modeling for fluid overload, we had things like age, severity of illness in there, we came up with a logistic regression model. And interestingly, in that regression model, medications were not significant contributors to that model and they weren't included in our final model. But we also did some machine learning analyses, which did perform a little bit better with regards to AUROC. So you can see here the random forest had the highest AUROC, or prediction value. And when we looked at the feature importance graph, which is a graph that can be plotted for different machine learning models to show what were the factors that contributed the most to that model's performance, and there were medications in those top features. And I think this is really important because we know medications are a causative agent for fluid overload. That's where the fluid is coming from, is IV medications, right? And so to have them not show up in a traditional prediction model is kind of odd. And you could say, well, sometimes just because it's a predictor doesn't necessarily mean that it's a causative agent, and that is definitely very true. But I think that when we're starting to get into trying to understand what are modifiable things, things that we can intervene upon as clinicians, I think it becomes more important to have our predictors reflect causative agents and things that we can actually change. So this was exciting to me to see that MRC showed up as a top feature within our fluid overload prediction models. We performed a very similar concept study from the fluid overload prediction model using prolonged mechanical ventilation. And again, we did a variety of regression modeling and then some different machine learning models. And we saw that the machine learning model did seem to have the highest AUROC, although not by a lot, which again I think speaks to maybe yet another interesting question about machine learning is, do we always need to be using it? And I don't know if that answer is necessarily true. But again, you could see that when we looked at the feature importance graph, medication showed up as top features that were contributing to those models. And so I think this kind of opens up the hypothesis that machine learning maybe is a very important methodology for medication-related questions, that it can handle more medication-related the complexity that's associated with that type of modeling. And I think that might be very important when we think about, again, thinking about clinicians and thinking about how we're going to optimize workload. These are kind of all pieces leading in that same vein. Another really interesting analysis by the MRC ICU investigator team evaluated pharmacophenotypes in the ICU. So the methodology here is unsupervised machine learning, which essentially just kind of looks at the data that you have and identifies clusters or patterns and asks, are these meaningful to the clinician? What was fascinating about this analysis was that we did identify six different phenotypes, or pharmacophenotypes, within our different patients of different types of medication use. And interestingly, those phenotypes did have significant differences with regard to outcomes. So just briefly explaining this picture, the first kind of few panels are showing how the data were processed. And then you can see in panel C, this is the unsupervised learning process. And then that resulted in the UMAP, or figure G, that shows the different medication clusters. And in the following slide, I'll show you how those clusters relate to outcomes. So this radar chart shows the overlap of different patient clusters with medication clusters and then the various clinical outcomes. And so what's interesting here is that you have something like patient cluster 5, which had the least serious outcomes, while patient clusters 1 and 4 have more severe outcomes. And then the question is, can you look at the different phenotypes and try to decide if there is any pattern within this? So what I found incredibly interesting about this paper and about this analysis is that when I looked at the clusters, I don't necessarily see them as, oh yeah, that makes total sense to me of why that cluster had more acute kidney injury because they had a bunch of medications that are associated with AKI. They were patterns that were not necessarily discernible by the human eye. And so I think that there's still a lot that we need to do with this to understand how do you apply this. It needs to be validated in a larger cohort and probably tighten up some of the clustering, maybe even have more groups to look at. But the concept that we need to be spending more time looking at medications regimens as a whole, how do they work in concert together, and how does that relate to outcomes? I think that's a pretty novel way to approach things because a lot of times when you see a study, it's like, okay, we're going to look at whether or not we should use vancomycin for MRSA. And it's yes, no to vancomycin, but we're not necessarily looking at the other 20 drugs that they were on and how those outcomes played into it. And the reason we haven't done that again is because all of those regimens are so different for all those patients and it just overwhelms our ability to do traditional regression modeling. So I think that this analysis supports the hypothesis that it's worth continuing to explore how to use these novel methods to understand more about medication use. One important component as we start to include medications into more of these analyses is thinking about a common data model. So a common data model is used for laboratory values, things like ICD-10, and there is one for medications as a whole called RxNorm which is put out by the National Library of Medicine. But there is a lot more to interpreting medications within a machine learning context or even just any type of data-driven context than necessarily drug and dose. And so this team has put out efforts to think through what are different features or different elements of a medication that describe it. So things like weight-based dosing, or root escalation, or the role of prophylaxis, how do you interpret a different dose and understand how to think through it? So an example that is fairly simple, but I still think is representative of some of the issues we face, is if I tell you that I was gonna take ibuprofen 200 and acetaminophen 325, you would say, okay, those are two starting doses, one tablet, and you wouldn't necessarily think anything of it. But a machine is potentially going to see 325 as more, numerically more than 200. So how do you help the machine understand that 325 of Tylenol is more or less equal to 200 of Advil? And these are interesting things to think through. Same again with thinking about the cefepime example I used earlier. How do you truly interpret the appropriate cefepime dose for that patient? Another key component that we have found is that MRC-ICU does relate to pharmacist workload. So again, we've seen correlations with kind of key process metrics of pharmacist interventions, order verification, eye events, as well as medication errors that have been caught by pharmacists. It's also important to realize the MRC-ICU has shown a stronger correlation to pharmacist interventions than a severity of illness indicator like SOFA score. It's important too to realize that although we are excited that these correlations do exist and that we do think that they're meaningful, they are moderate in strength and I think are more so indicators than necessarily a pure reflection. So one question is, do we need other indicators of workload entirely? And I'll also posit maybe we need a different metric entirely. So I'll challenge the audience to come up with something better. And I'm excited to see what that is. However, I think key takeaways are that medications are related to outcomes and we can see that through the score and that those managing medications do affect outcomes. And I think that this score helps link those key concepts all together and that is really important. So in bringing all of these concepts together, ideally the goal is to have data-driven workload optimization, that we would take complex ICU data, labs, vital signs, medications, severity of illness, diagnosis, all of that stuff, funnel it through machine learning concepts and maybe funnel it through a metric like MRC ICU or even some type of computer program that then is going to output clinically relevant medication predictions. And these predictions to me have, the world is our oyster a little bit. These could be predictions like, hey, you're gonna have 50 interventions on this patient over the next two days. So this needs to be your pharmacist to patient ratio. But it could also be saying, hey, this patient's at high risk for fluid overload or mechanical ventilation. You might wanna take a second look and see if there's anything that we can do to modify. So the key takeaway of this section is that medications are important predictors of patient outcomes and artificial intelligence plus metrics has the potential to be used for workload and outcome prediction. So with all this being said, I wanna try to bring it together to what we think some best practices are and how can we implement them. When I think about what a critical care pharmacist offers, to me, pharmacists support a culture of evidence-based medication use. This is a culture that permeates the whole ICU, the whole team that is there. And importantly, that culture can be established by just the presence of that person and by all those kind of like little things that are hard to quantify. But it's the fact that you gave the little nod or gave the little smile to somebody when they said something or you sent the email with the newest paper or you gave an in-service or all of those things come together to support a culture that is trying to optimize medication benefits and reduce harms. When I think about this culture, I think about a triple domain of critical care pharmacist value. And this domain is direct patient care that is going on rounds, making interventions, looking at the chart and stuff along those lines. But it's also indirect patient care. This is the quality improvement stuff. This is guidelines, it's protocols, it's order sets, it's medication use evaluations, it's sitting on a root cause analysis team, it's sitting on various committees. And those things I think are incredibly important and very, very valuable to have a pharmacist perspective on there. And I think should be valued as ways that we are improving medication-related outcomes and also patient outcomes. And finally, through professional service. Professional service here, I mean a broad category, things like doing research, things like being involved with education in the PharmD curriculum or precepting, things like sitting in national organizations. So we have a critical care pharmacist sitting on practice guidelines for ARDS or sepsis and improving how those guidelines are written and thinking through how to optimize the medication components of that. Even just little things of, as opposed to writing hydrocortisone 200 milligrams a day, which is how it was written in the original surviving sepsis guidelines, versus actually stating that it's 50 milligrams Q6 or continuous infusion. Those details matter. And that actually is a real story. I had a physician for many years who told me that it didn't matter, the dosing, as long as it was 200 a day. So it could be 200 once a day or 100 twice a day. And I was like, that's not how it works or how it was studied. But those types of things, it's now corrected in the guidelines because we had pharmacists that were involved in that. So all of these things come together to improve how we give medications and how we use medications in the ICU. And I think that's incredibly important for us to recognize. There do need to be some paradigm shifts for pharmacist team integration. So I think there are some historical views that we need to change and to think through how we're gonna be seen differently and how we approach a team with these things in mind. So historically, it was pharmacists provide a commodity. We provide a drug, as opposed to thinking of us as providing cognitive services. We were responsible for discrete tasks. We did vancomycin PK, or we did transitions of care, as opposed to seeing us as responsible for the patient. And I think that's similar to our nursing and physician colleagues, who are taking more of a global responsibility. Historically, we looked at workload as being measured by a frequency of tasks. So how many interventions? How many orders? As opposed to looking at the intensity of the patient needs. Because the thing is, a frequency of a task does not capture the cognitive load of reviewing a patient. So sometimes you get a patient who's on heparin and vancomycin and an aminoglycoside and three different continuous infusions for sedation and multiple vasopressors, but you don't have anything to change. But that doesn't mean you didn't look at all of those things and all of the labs and all of the vital signs and the ins and the outs to come up with the decision that that is good and that is what the patient should be on. So that's the intensity of the patient needs that is, again, maybe not captured by making a change. But it doesn't mean that because no change was made that someone shouldn't look at it. We used to have a lot of stuff based on error catching and safety of administering meds in the ICU. And the great part about barcoding and different types of CPOE is we do have fewer errors of that nature. Although I will say that my team conducted an evaluation of errors and we found like 600 errors in an eight week period by just two pharmacists. So even in the modern day, we still have a fair amount of errors. But as opposed to purely error catching, we're looking at optimization of the choices. So the choice, again, I love using piperacillin, tezobactam versus cefepime. That might not be wrong to use one agent versus another, but we might say that one is optimal or optimized. And that we think that those things contribute to outcomes. Again, thinking about medication changing occurring as a reactive intervention versus proactive decision making. You're at the table for all of these types of conversations. And again, that optimal outcomes are gonna be achieved not by an individual professional, but by the systems of clinician care that are there. So these are things that I think are paradigms we're moving towards and that we need to continue to move towards. Given this concept of a culture of evidence-based medication use, I think we need to recognize that everything we've got is an imperfect measurement. I love these two quotes, and I love that they're almost saying the opposite thing. So the first says, what gets measured gets improved. To some extent, if we cannot quantify what we're doing, then it's going to be very hard for us to improve anything. At the same time, when you take a metric and it becomes the target, it ceases to be a good metric. So it's not that intervention tracking is bad. It maybe is very useful in a time-delimited fashion to observe some particular component of something. But it also shouldn't be your target. Because again, thinking about my neuromuscular blockade protocol, I actually, by making that guideline and order set, I reduced the number of interventions that I made. But I think we would all agree that I improved care by doing that. So the metric of interventions is not a good target. And there are lots of metrics that aren't good targets. So we need to think through why are we measuring it and what is that doing? So again, there are these intangibles, things like promoting evidence-based culture, cognitive services, that I don't think we're ever gonna be able to perfectly measure. But I do think metrics are a tool and something that we can think through. And so the next couple of slides, I wanna talk about how we can maybe link some of these features into meaningful things and then maybe measure those as a way to truly improve outcomes and improve workload for that matter. So a lot of the goal of what we've been discussing is how can we link patient features to workload? So it'd be nice to be able to take patient features that can be used as predictors, something like SOFA, maybe something like MRC-ICU, and that that's gonna then predict quantifiable workload characteristics, something like the pharmacist-to-patient ratio. And from there, this is going to then feed into the patient-centered outcome of interest, like mortality. So this ability to link patient features to prediction of pharmacist workload, as well as outcomes, and then linking that workload to outcome, I think those are areas of study. And I think that we're getting there, and we're certainly starting to think about these questions. But again, my challenge to all of you, and particularly as we get into the next slide, is to think about how we can better connect these components and think about what are ways that we can try to characterize the amount of work that a pharmacist is doing based on patient characteristics, and then link that to metrics that are meaningful. So a potential concept is something I've been calling a clinician-designed dashboard. So your dashboard would ideally reflect these various domains of value that support the culture of evidence-based medication use. So number in services, MUEs, learners precepted, leadership positions, all of that is really meaningful. But then I think what would be really neat is to think about quality indicators, or things that are agreed upon by the team. So maybe at your specific institution, there's a real problem with using double anaerobic coverage. And you say, hey, we're gonna really work on this. So we're gonna put that on our dashboard, and we're gonna focus on getting rid of all the extra flagell. Or you say, we've really had an issue with too much deep sedation. We're gonna put that as our metric, and we're gonna focus on that. I think the ability to say that the pharmacists are gonna take on some type of goal, measure it, and see what they can do to nudge the needle is really very meaningful. Again, this is not pharmacists taking sole responsibility for a RAS score, or inappropriate antibiotics. But it is saying, these are things that we think are important, and that we're trying to work towards. And I think the ability to think in these terms of how we are improving the general quality of care, and how we're working together with our ICU team towards something. I wish we did more of this, and could look at a graphic like this to show all the wonderful things that a pharmacist does. So, what can we do today? This is a toolkit. It's got about five different steps in it for improving patient access to pharmacists. One of the clearest things is if there's a multi-professional rounding team that doesn't have an ICU pharmacist, that is a really big area to try to justify that position. And again, that's based on the Lee meta-analysis that's really showing that to not have that, you are increasing the risk of mortality, longer length of stay, as well as adverse drug events. I think institutional privileging is an important component to improving access to pharmacist services. It can streamline so much. So the ability to order labs, manage different components of ICU medication profiles, I think there's a lot of value in that. And if nothing else, it just makes your day more efficient. I also think that pharmacists should be on the different healthcare and safety teams within the institution. So if there's a committee that a pharmacist isn't on, I think there should be one on there. There is too many things that are medication-related in the service of patients that it requires that expertise. I do think that there are opportunities to think about how we can expand critical care pharmacy residency training programs, medication pass-through funding, organizational grants, so forth, come to mind. And finally, thinking about how we can evaluate and document existing workload of pharmacists. So is there a way to think about the number of patients one has taken care of on average, think about the rate of cross-coverage, think about the rate of how many patients, how often do they get rounded on? The reality is if a patient comes in on Friday, they will spend the first two-thirds of their ICU stay without a pharmacist rounding in a vast majority of situations. And personally, I find that unfortunate. I would love to say that the answer is this sweeping overhaul that we're gonna make big changes, but likely it's gonna be a stepwise progression for workload optimization. So you could think about things ranging from decentralized services or tele-ICU to establishing a pharmacist per ICU to establishing a pharmacist per rounding team to thinking about flex positions to reduce cross-coverage, to think about adding evening services to ultimately, ideally, going 24-7, 365. 24-7, 365 with these type of services. And to me, I think this is where we need to go ultimately. I think that what pharmacists offer is just way too important for us to only be there day shift, weekdays, with no holidays and stuff like that. So these are things I would like to see us move towards and you can kind of see like, okay, what do you think is the next incremental step? There is a study currently ongoing called the Optimizing Pharmacist Team Integration for ICU Patient Management, or Optum. This study is trying to understand the patient-level pharmacist-to-patient ratio. So what was the average number for the patient for that workload? So we're trying to see here that as a patient was one of more patients that a pharmacist was taking care of that potentially they had worse outcomes with the goal that we can maybe establish a pharmacist-to-patient ratio. This has been quite a large undertaking as of this recording. We're at 20,000 patients that have been enrolled, trying to see the outcomes here, mortality, length of stay, ventilator-free days, so forth. I think that one of the most interesting components of this is that we have a fair amount of data that's gonna be captured about physicians, advanced practice providers, nurses, as well as pharmacists, learners, trainees. And I'm curious to be able to see how more well-resourced ICUs are potentially providing better care with the idea that maybe there are certain teams that have more people or the right combination of people that are associated with better outcomes. So stay tuned for this. Very exciting to see where this is leading. A final concept that's going to be explored in the Optum study that I am very excited about is a novel workload index. This is a combination of what we're calling a patient burden index and a clinical burden index, which is looking at, again, number of different patients one is taking care of, but then also these other factors, number of medical teams, location, typical patient population, trying to come up with a better idea of what workload really means in relation to outcomes. So this is, again, novel. It hasn't been studied before. We're hoping that it makes sense based on previous literature. And again, stay tuned for how we can maybe use things like this to do a better job with workload optimization. So the key takeaway from this section is that best practices are gonna be those that integrate pharmacists into the ICU healthcare team and support the development of a culture of evidence-based ICU medication use. In conclusion, pharmacists are essential members of the healthcare team. Workload is an important quality metric for ICU patient outcomes for all clinicians. Drugs are independent risk factors for outcomes that require expert management, likely from a pharmacist. And the best practices are gonna keep clinicians at the bedside with the patient and with the team. And again, I think these are investigations that are obviously quite relevant for critical care pharmacists, but I think have ramifications for how we approach the care of an ICU patient no matter what the profession is and as a team. Thank you all for listening to today's presentation. It was an absolute pleasure talking about one of my favorite topics. Looking forward to being able to take any questions that you all may have. You can connect with me at andrea.sakora or my email is sakora.uga.edu. Always happy to chat more on any of these things.

ICU pharmacists

workload optimization

patient outcomes

clinician workload

pharmacist practice models

medication use

pharmacist team integration

patient management

evidence-based care